This comes from a discussion in this issue:
https://issues.dlang.org/show_bug.cgi?id=13122

Perhaps this topic was already discussed and solved in past, but unfortunately I don't remember such discussions.

Issue 13122 is about cartesianProduct but the same situation is visible more simply with std.algorithm.filter:

void main() {
    import std.algorithm: filter;
    immutable a = "ABC";
    foreach (t1; a.filter!(x => true))
        pragma(msg, typeof(t1));
    immutable b = "ABC"d;
    foreach (t2; b.filter!(x => true))
        pragma(msg, typeof(t2));
}


Output:

dchar
immutable(dchar)


So cartesianProduct seems to be working like the other Phobos functions and I'll have to close Issue 13122 as worksforme.


a.filter yields dchar because range iteration converts ranges of chars to ranges of dchars. But then why isn't the output like this?

immutable(dchar)
immutable(dchar)

The string a is composed of immutable(char) after all.

For me it's strange that iterating on arrays of immutable chars or immutable dchars gives an output of different "constness". This looks like a gratuitous breakage of symmetry :-)

Bye,
bearophile